dreamforge
Turn world-model imagination into corrective training data for robot policies — a single-GPU, falsifiable pipeline (SmolVLA · LIBERO · CEM · paired simulator replays).
- embodied-ai
- imitation-learning
- lerobot
- libero
- pytorch
- research
- robotics
- vision-language-action
- vla
- world-model
- Stars
- 123
- Forks
- 3
- + today
- +5
- Created
- 2d
Ranking data as of October 13, 2026 (UTC).
Star History
Today, hour by hour
Overview
Turn world-model imagination into corrective training data for robot policies — a single-GPU, falsifiable pipeline (SmolVLA · LIBERO · CEM · paired simulator replays). It ranks #47 on GitTiger, gaining +5 stars on October 13, 2026 (UTC).
The project is written in Python and has 3 forks. It was created 2d ago.
What it is
dreamforge is a Python research pipeline described as turning world-model imagination into corrective training data for robot policies. The description calls it a single-GPU, falsifiable pipeline using SmolVLA, LIBERO, CEM, and paired simulator replays. It is MIT licensed.
Capabilities described
- Corrections are replayed in the simulator, and model-guided results are compared against a budget-matched baseline (README preview). - Statistics include paired exact McNemar tests and Wilson intervals (README preview). - The preview mentions pipeline bugs found during testing and a checklist item to replicate a targeted-training result across seeds. - Topics include imitation-learning, vision-language-action, and world-model.
Getting started
No installation steps were supplied. See the README at https://github.com/rzc156/dreamforge.
git clone https://github.com/rzc156/dreamforge.git
cd dreamforge
# see README for setupFrequently Asked Questions
What problem does it address?
Per the description, it converts world-model imagination into corrective training data for robot policies.
How are results checked?
The preview says corrections are replayed in the simulator and compared to a baseline using paired McNemar tests and Wilson intervals.
What environment does it need?
The description says single-GPU and names SmolVLA, LIBERO, and CEM. Detailed requirements were not supplied.